Re: Streams in Caml Light 0.5

Michel Mauny (Michel.Mauny@inria.fr)
Mon, 26 Oct 92 17:10:11 MET

Subject: Re: Streams in Caml Light 0.5
To: HEDDEN@ESDSDF.dnet.ge.com
Date: Mon, 26 Oct 92 17:10:11 MET
In-Reply-To: <9210260033.AA01662@aitgw.ge.com>; from "HEDDEN@ESDSDF.dnet.ge.com" at Oct 25, 92 7:33 pm
From: Michel.Mauny@inria.fr (Michel Mauny)

> I am having a small difficulty using streams in Caml Light v0.5.
>
> The example below might be a bug, but is more likely due to a
> lack of understanding on my part.
>
> The exception in the following toplevel session indicates that
> there is a problem:
>
>
> > Caml Light version 0.5
>
> #include "test";;
> map_stream : ('a -> 'b) -> 'a stream -> 'b stream = <fun>
> filter_stream : ('a -> bool) -> 'a stream -> 'a stream = <fun>
> int_stream : int stream = <abstract>
> - : int = 1
> - : int = 9
> - : int = 25
> odd_sq_stream : int stream = <abstract>
> - : int = 49
> Uncaught exception: Parse_failure
> - : unit = ()
> #quit();;

The problem is that your map_stream and filter_stream are incorrect
(if I understand correctly what they should do) they both return a
stream with only one element. Your int_stream (correctly) contains
the stream of strictly positive integers. Considering that a
successive match on a stream *physically* destroys the stream. Now,
what you do is select the first odd element (1), compute its square,
and return a stream with that single element. The next time you do
it, it returns 9 and so on. These calls succeed because filter_stream
builds a new stream (with only one element) each time Now, if you
encapsulate the process into a function, map_stream empties the result
of filter_stream after the first call, and fails (the stream is empty)
for subsequent calls. Cf. the comments and corrections below.

> The file "test.ml" (that is loaded above) contains the following:
>
>
> (* Maps a function over a stream *)
> let rec map_stream func = function
> [< 'x >] -> [< 'func x >]
> | [< >] -> [< >]
> ;;

I guess your map_stream function is supposed to build the stream of
results of applying a function `func' on *all* elements of the
argument stream. In that case, the correct definition should be:

let rec map_stream func = function
[< 'x ; (map_stream func) str >] -> [< '(func x); str >]
| [< >] -> [< >]
;;

You missed the recursive call.

> (* Filters a stream according to a predicate *)
> let rec filter_stream pred strm =
> match strm with
> [< 'x >] -> if pred x then [< 'x >]
> else filter_stream pred strm
> | [< >] -> [< >]
> ;;

Now, filter_stream is also wrong, for the same reason (a resursive
call is missing). The correct definition should be:

let rec filter_stream pred = function
[< 'x ; (filter_stream pred) str >]
-> if pred x then [< 'x ; str >] else str
| [< >] -> [< >]
;;

For streams, it is very common that recursive calls occur in the
`stream pattern' (this is by analogy with BNF grammars, cf. the part
of the documentation dedicated to streams and parsers).

> (* The stream of positive integers *)
> let int_stream = ints 1
> where rec ints n = [< 'n; ints (succ n) >]
> ;;
>
> (* Use `filter_stream' and `map_stream' to create a
> stream of the squares of the odd integers *)
> (* These work correctly *)
> stream_next (map_stream (function x->x*x)
> (filter_stream (function n->(n mod 2)= 1)
> int_stream));;
> stream_next (map_stream (function x->x*x)
> (filter_stream (function n->(n mod 2)= 1)
> int_stream));;
> stream_next (map_stream (function x->x*x)
> (filter_stream (function n->(n mod 2)= 1)
> int_stream));;

Above, all calls succeeded because filter_stream builds a new
`singleton' stream, map_stream returns another singleton stream, and
you requested only for its first element.

> (* Encapsulate the above *)
> let odd_sq_stream =
> map_stream (function x->x*x)
> (filter_stream (function n->(n mod 2)= 1)
> int_stream)
> ;;
>
> (* Test the above encapsulation *)
> (* This works correctly *)
> stream_next odd_sq_stream;;
>
> (* This causes an error *)
> stream_next odd_sq_stream;;

The encapsulation `odd_sq_stream' of the stream contains only one
element, this is why the second call failed.

Best regards,

Michel
----------------------------------------------
INRIA -- BP 105 -- F-78153 Le Chesnay Cedex
Tel.: +33 1 39 63 57 96 Fax: +33 1 39 63 53 30
Email: Michel.Mauny@inria.fr
----------------------------------------------